Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / The Hidden Cost of Prototyping Iterations
Transcript
- Lucas: There's this number that I cannot stop thinking about. A medical-device startup I've been following — they went through nine prototype revisions before they had a single working unit. Nine. And the total bill for those revisions? Three hundred and forty thousand dollars. Luna: Ouch. And I'm guessing that's before they even got to regulatory testing or anything clinical. Lucas: Exactly. Pre-clinical, pre-anything. Just getting the physical thing to function the way the CAD said it should. And this is not unusual. There's a rule of thumb in hardware circles — each prototype revision costs roughly two and a half times the previous one. So your first revision might be twenty thousand, the next fifty, the next a hundred and twenty-five. By revision four you're past three hundred grand. Luna: And the thing is, most founders don't budget for that. They budget for one or two revisions. The pitch deck says 'we'll need eighty thousand for prototyping.' Eight months later they've burned three-fifty and the board is asking questions. Lucas: Right. So the question I want to drill into today is — how do you break that cycle? Because the obvious answer is 'iterate faster,' but that's vague. What does faster iteration actually look like when you're dealing with physical objects, supply chains, and machine shops that have six-week lead times? Luna: I'd love to hear what the smart founders are doing differently. Because the story you just told — nine revisions, three-forty K — that sounds like a failure of strategy, not just bad luck. Lucas: I think it is. And the shift I'm seeing is from what I'd call 'iterative prototyping' to something I've heard called 'stepping-stone prototyping.' The difference is that iterative prototyping assumes each version gets you closer to the final product. Stepping-stone prototyping deliberately builds something you know you'll throw away — but you build it fast and cheap, just to test one specific assumption. Luna: So you're not trying to make the thing. You're trying to kill the riskiest unknown first. Lucas: Exactly. The medical-device startup I mentioned — they were trying to make the final device from revision one. Every iteration they were adding features, refining tolerances, chasing the real thing. A stepping-stone approach would have said: what's the single hardest thing to get right? For them it was the fluidics — the precise delivery of a very small dose. So the first prototype should have been just a fluidics test on a breadboard, not a full device housing. Luna: That makes so much sense. And it's funny because software startups learned this lesson years ago — minimum viable product, build the smallest thing that tests the riskiest assumption. But hardware founders seem to forget it the minute they start talking to injection molding shops. Lucas: Part of that is cultural. Hardware people are trained to build things properly. There's a pride in craftsmanship. The idea of building something deliberately shoddy feels wrong. But the economics are brutal — the cost of over-building early is orders of magnitude higher than the cost of building a quick and dirty test. Luna: What about tools? I know there's been a wave of desktop manufacturing equipment that's supposed to make early prototyping cheaper. Is that actually changing the math? Lucas: It is, but not in the way the marketing suggests. Take the Bantam Tools Desktop PCB Milling Machine. It's about three thousand dollars. It lets you mill circuit boards in-house instead of waiting two weeks for a fab house. That saves time and money on board revisions — but it doesn't solve the mechanical prototyping problem. You still need enclosures, moving parts, heat management. Those are the expensive iterations. Luna: So desktop tools help on the electrical side, but the mechanical side still requires traditional methods — unless you're doing something like 3D printing. Lucas: 3D printing definitely helps, especially for form-factor testing. But the materials are usually not production-grade. So you still need a machined prototype for anything that's going to bear load or handle heat. That's where the cost really lives — in the precision machining. Luna: Let's talk about that medical-device startup again. What did they actually end up doing after the nine revisions? Did they get a working device? Lucas: They did, but only after they changed their approach entirely. They stopped trying to make the final device and instead sent a simplified version to a precision machine shop in Shenzhen. That shop did the entire assembly in three weeks for forty thousand dollars — about a tenth of what they'd spent on their last domestic revision. Luna: Wow. So the lesson wasn't just 'iterate smarter,' it was also 'find the right manufacturing partner earlier.' Lucas: Absolutely. And that's a whole other dimension. A lot of founders treat manufacturing as something you figure out after you have a finalized design. But the smart ones bring a manufacturing engineer into the process from prototype one. They design for manufacturability from the start, which means fewer revisions because you're not going back to fix things the factory can't actually do. Luna: There's a great example from the robotics world — the company that makes the 'Stretch' warehouse robot. I read that they went through twenty-seven major design iterations before they shipped a single unit. But they kept each iteration cheap by using off-the-shelf components and doing most of the assembly in-house. Lucas: Twenty-seven iterations sounds insane, but if each one cost five thousand instead of fifty thousand, that's actually a reasonable R&D spend. The key metric isn't number of iterations — it's cost per iteration. The best hardware startups optimize for cheap iterations, not few iterations. Luna: Right. And that's where the stepping-stone mentality really pays off. You design each prototype to answer one question as cheaply as possible. If you need to test thermal performance, you 3D print a rough housing and glue on a heat sink. You don't cnc machine a beautiful aluminum enclosure. Lucas: By the way, if today's episode was useful to you and you want to keep it ad-free, buy me a coffee dot com slash fexingo helps. That's literally what funds these conversations — a handful of listeners who chip in a few bucks a month. Luna: Yeah, and it means we can keep digging into real numbers and real case studies without any sponsor influence. So if that matters to you, the link is buy me a coffee dot com slash fexingo. Lucas: So back to that medical-device startup. After they switched to the Shenzhen shop, they got their first working prototype in three weeks. Then they did two more revisions — each one costing about fifteen thousand — and they had a production-ready design. Total prototyping spend after the pivot: seventy thousand. Total before the pivot: three-forty. So they saved two hundred and seventy thousand dollars just by changing their approach. Luna: That's a massive difference. And I imagine that made their pitch to investors a lot stronger too. Instead of saying 'we need another five hundred thousand for more prototypes,' they could say 'we have a working unit and we need capital for tooling and inventory.' Lucas: Exactly. And that's something I'm seeing more VCs pay attention to. I've talked to a few early-stage hardware investors who now ask for a 'prototype budget' in the pitch deck — not just the total amount, but a breakdown of how many iterations, at what cost per iteration, and what assumptions each one tests. If a founder can't articulate that, it's a red flag. Luna: That should honestly be standard practice. Because the founders who can answer that question are the ones who understand the physics of their own business. Lucas: And honestly, the best hardware founders I've met treat prototyping the way a scientist treats an experiment. They have a hypothesis, they design a test, they run it, they learn, they iterate. They don't get emotionally attached to the prototype itself. And that's hard — because when you've spent months designing something, it's painful to admit it's wrong. But the ones who can do that are the ones who survive. Luna: I think a lot of listeners are probably in the middle of a prototype cycle right now. What's the one thing you'd tell them to do differently starting tomorrow? Lucas: Identify the single most expensive assumption in your design — the thing that, if you get it wrong, forces a complete redesign — and build the cheapest possible test for that assumption. If it's weight, build a crude mockup and put it on a scale. If it's battery life, wire up a development board and run a discharge test. Do not touch the final enclosure until you've validated that assumption. That one shift will save you at least two revision cycles. Luna: That's great advice. And it's the kind of thing that sounds obvious once you hear it, but in the heat of development, it's easy to skip. Lucas: It is. And that's why conversations like this matter — because the hard part isn't the technology. It's the discipline to slow down and test the right thing at the right time. Luna: Alright, so for next time — I want to talk about something I saw in a recent hardware incubator demo day. A startup that's using a novel approach to supply chain financing. Think there's something there? Lucas: I think there's a lot there. Let's dig into it next episode.